CN110121883A

CN110121883A - The method and apparatus that image is decoded in image encoding system

Info

Publication number: CN110121883A
Application number: CN201680091905.1A
Authority: CN
Inventors: 徐廷东; 林宰显
Original assignee: LG Electronics Inc
Current assignee: LG Electronics Inc
Priority date: 2016-12-05
Filing date: 2016-12-05
Publication date: 2019-08-13
Anticipated expiration: 2036-12-05
Also published as: CN110121883B; KR102414924B1; US20220264141A1; PT3855745T; HUE055198T2; CN116527884A; DK3541073T3; KR20190084271A; ES2886871T3; US20200366922A1; CN116527885A; KR20210135347A; US10764598B2; EP3855745A1; EP3855745B1; PL3855745T3; PL3541073T3; KR20220093392A; WO2018105757A1; HUE059684T2

Abstract

The picture decoding method according to the present invention executed by decoding apparatus is the following steps are included: obtain the information of the inter-prediction about object block by bit stream；Based on the information about inter-prediction come the motion information of derived object block；Forecast sample is derived by executing the inter-prediction for object block based on motion information；Reconstructed blocks are generated based on forecast sample；Based on reconstructed blocks come the modified motion information of derived object block；And the motion information of object block is updated based on modified motion information.According to the present invention, the modified motion information of object block can be calculated, and the motion information of object block can be updated based on modified motion information, so that overall coding efficiency can improve.

Description

The method and apparatus that image is decoded in image encoding system

Technical field

The present invention relates to the technologies encoded for image, more particularly, to one kind to image in image encoding system The method and apparatus being decoded.

Background technique

To high-resolution, high quality graphic (for example, (superelevation is clear by HD (fine definition) image and UHD in various fields Degree) image) demand be continuously increased.Since image data has high-resolution and high quality, so relative to traditional images number According to the amount for the information or bit to be sent increases.Therefore, it is sent out when using such as traditional wire/WiMAX route medium Send image data or using existing storage medium come when storing image data, transmission cost and carrying cost increase.

Therefore, it is necessary to a kind of efficient image compress techniques effectively to send, store and reproducing high resolution and high quality The information of image.

Summary of the invention

Technical problem

The present invention provides a kind of methods and apparatus for improving image coding efficiency.

The present invention also provides a kind of methods and apparatus for inter-prediction, update the motion information of object block.

The present invention also provides a kind of for calculating the modified movement letter of object block after the decoding process of object block It ceases and is based on modified motion information and carry out more new method and apparatus.

The present invention also provides a kind of updates that object block is used for next piece of the motion information neighbouring with object block Motion information method and apparatus.

Technical solution

On the one hand, a kind of picture decoding method executed by decoding device is provided.The picture decoding method include with Lower step: the information of the inter-prediction about object block is obtained by bit stream；It is pushed away based on the information about inter-prediction Lead the motion information of object block；Forecast sample is derived by executing the inter-prediction for object block based on motion information；Base Reconstructed blocks are generated in forecast sample；Based on reconstructed blocks come the modified motion information of derived object block；And based on modified Motion information updates the motion information of object block.

On the other hand, it provides a kind of for executing the decoding device of image decoding.The decoding device includes: entropy decoding Device is configured as obtaining the information of the inter-prediction about object block by bit stream；Fallout predictor is configured as being based on Information about inter-prediction carrys out the motion information of derived object block, and the interframe of object block is directed to by executing based on motion information Prediction generates the reconstructed blocks of object block based on forecast sample, and based on reconstructed blocks come derived object to derive forecast sample The modified motion information of block；And memory, it is configured as updating the movement of object block based on modified motion information Information.

On the other hand, a kind of method for video coding executed by encoding device is provided.Method includes the following steps: Generate the motion information of object block；Forecast sample is derived by executing the inter-prediction for object block based on motion information； Reconstructed blocks are generated based on forecast sample；The modified motion information of object block is generated based on reconstructed blocks；And based on amendment Motion information update the motion information of object block.

On the other hand, a kind of video encoder is provided.The encoding device includes: predicting unit, is configured as The motion information for generating object block derives forecast sample by executing the inter-prediction for object block based on motion information, Reconstructed blocks are generated based on forecast sample, and the modified motion information of object block is generated based on reconstructed blocks；And storage Device is configured as updating the motion information of object block based on modified motion information.

Beneficial effect

According to the present invention, after the decoding process of object block, the modified motion information of object block is calculated, and can be more It is newly more accurate motion information, overall coding efficiency can improve as a result,.

According to the present invention, next piece of the motion information neighbouring with object block can be based on the motion information of the update of object block It derives, and the propagation of distortion can be reduced, as a result, overall coding efficiency can improve.

Detailed description of the invention

Fig. 1 is the schematic diagram for showing the configuration for the video coding apparatus that the disclosure is suitable for.

Fig. 2 is the schematic diagram for showing the configuration for the video decoder that the disclosure is suitable for.

Fig. 3 shows the case where executing inter-prediction based on one-way movement information and based on the two-way fortune for being applied to object block The example for the case where dynamic information executes inter-prediction.

Fig. 4 shows the example of the cataloged procedure of the method for the motion information including updating object block.

Fig. 5 shows the example of the decoding process of the method for the motion information including updating object block.

Fig. 6 shows the example for updating the method for motion information of object block by block matching method by decoding device.

Fig. 7 shows the reference pictures that can be used in modified motion information.

Fig. 8 show in the case where the motion information of object block is updated to the modified motion information of object block with target The example of next piece neighbouring of merging candidate list of block.

Fig. 9 show in the case where the motion information of object block is updated to the modified motion information of object block with target The example of next piece neighbouring of merging candidate list of block.

Figure 10 show in the case where updating the motion information of object block based on the modified motion information of object block with The example of next piece neighbouring of motion vector predictor candidate list of object block.

Figure 11 show in the case where updating the motion information of object block based on the modified motion information of object block with The example of next piece neighbouring of motion vector predictor candidate list of object block.

Figure 12 schematically shows the method for video coding of encoding device according to the present invention.

Figure 13 schematically shows the video encoding/decoding method of decoding device according to the present invention.

Specific embodiment

The disclosure can be modified according to various forms, and particular implementation will be described in the accompanying drawings and show.However, this A little embodiments are not intended to limit the disclosure.Term used in being described below is only used for description particular implementation, and It is not intended to be limiting the disclosure.Odd number expression includes plural number expression, as long as its is interpreted significantly differently.Such as " comprising " and " having " Term be intended to refer to the presence of the feature used in being described below, quantity, step, operation, element, component or combinations thereof, because This it should be understood that do not preclude the presence or addition of one or more different characteristics, quantity, step, operation, element, component or its A possibility that combination.

On the other hand, the element in attached drawing described in the disclosure for convenience of explanation different specific function and it is only It on the spot draws, is not meant as element and is implemented by separate hardware or stand alone software.For example, two or more yuan in element Part can be combined to form discrete component or an element may be logically divided into multiple element.In the design for not departing from the disclosure In the case of, the embodiment that element is combined and/or divides belongs to the disclosure.

Hereinafter, will be described in detail with reference to the accompanying drawings embodiment of the present disclosure.In addition, running through attached drawing, similar label is used for Indicate similar element, and the identical description by omission about similar components.

In the present specification, in general, picture refers to that the unit for indicating the image of specific time, slice are constitute picture one Partial unit.One picture can be made of multiple slices, and term picture and slice can be mixed with each other when necessary.

Pixel can refer to constitute the minimum unit of a picture (or image).In addition, " sample " can be used as it is corresponding with pixel Term.Sample can usually indicate the value of pixel or pixel, can only indicate the pixel (pixel value) of luminance component, and can only table Show the pixel (pixel value) of chromatic component.

The basic unit of unit instruction image procossing.Unit may include in specific region and information related with the region At least one.Optionally, unit can be with the term hybrid in block, region etc..In a typical case, M × N block can the row of expression Arrange into M column and the sample of N row or the set of transformation coefficient.

Fig. 1 schematically illustrates the structure for the video coding apparatus that the disclosure is suitable for.

Referring to Fig.1, video coding apparatus 100 includes image splitter 105, fallout predictor 110, subtracter 115, converter 120, quantizer 125, rearrangement device 130, entropy coder 135, inverse quantizer 140, inverse converter 145, adder 150, filter 255 and memory 160.

Input picture can be split at least one processing unit by image splitter 105.Here, processing unit can be volume Code unit (CU), predicting unit (PU) or converter unit (TU).Coding unit is the cell block of coding, maximum coding unit (LCU) coding unit of depth can be split into according to quad-tree structure.In this case, maximum coding unit is available Make final coding unit or coding unit can be recursively split into as needed depth coding unit and can root The coding unit with optimal size is used as by final coding unit based on code efficiency according to video properties.When setting minimum code When unit (SCU), coding unit can not be split into coding unit more smaller than minimum coding unit.Here, final coding is single Member refers to the coding unit for being divided or being split into fallout predictor or converter.Predicting unit is the block divided from coding unit block, And it can be the cell block of sample predictions.Here, predicting unit is divided into sub-block.Transform block can be according to quad-tree structure From coding unit block splitting, and it can be the cell block for deriving transformation coefficient and/or derive residual signals from transformation coefficient Cell block.

Hereinafter, coding unit is referred to alternatively as encoding block (CB), predicting unit is referred to alternatively as prediction block (PB), and converter unit can Referred to as transform block (TB).

Prediction block or predicting unit can refer to have block-shaped specific region in picture, and may include forecast sample Array.It in addition, transform block or converter unit can refer to have block-shaped specific region in picture, and may include transformation coefficient Or the array of residual sample.

Fallout predictor 110 can execute prediction to processing target block (hereinafter, current block), and produce pre- including current block The prediction block of test sample sheet.The unit of the prediction executed in fallout predictor 110 can be encoding block, perhaps can be transform block or It can be prediction block.

Fallout predictor 110 can be determined whether to current block application intra prediction or using inter-prediction.For example, fallout predictor 110 It can be determined whether as unit of CU using intra prediction or inter-prediction.

In the case where intra prediction, fallout predictor 110 can based in the picture (hereinafter, current picture) belonging to current block when The baseline sample in preceding piece of outside derives the forecast sample of current block.In this case, fallout predictor 110 can be based on current block It is derived forecast sample (situation (i)) adjacent to the average or interpolation of baseline sample, or can be based on the neighbouring benchmark sample of current block Should have in the baseline sample that is present on specific (prediction) direction about forecast sample derive forecast sample (situation (ii)).Feelings Condition (i) is referred to alternatively as non-directional mode or non-angled mode, and situation (ii) is referred to alternatively as directional pattern or angle mode.In frame In prediction, as an example, prediction mode may include 33 directional patterns and at least two non-directional modes.Non-directional mode can wrap Include DC mode and plane mode.The prediction mode applied to contiguous block can be used to determine current block to be applied in fallout predictor 110 Prediction mode.

In the case where inter-prediction, fallout predictor 110 can based on sample specified by the motion vector on reference pictures come Derive the forecast sample of current block.Fallout predictor 110 can be by applying skip mode, merging patterns and motion-vector prediction (MVP) Any of mode derives the forecast sample of current block.In the case where skip mode and merging patterns, fallout predictor 110 can Use the motion information of contiguous block as the motion information of current block.It is different under merging patterns in the case where skip mode, The difference (residual error) between forecast sample and original sample is not sent.In the case where MVP mode, the motion vector of contiguous block is used as Motion vector predictor accordingly acts as the motion vector predictor of current block to derive the motion vector of current block.

In the case where inter-prediction, contiguous block may include the spatial neighbor block being present in current picture and be present in base Time contiguous block in quasi- picture.Reference pictures including time contiguous block are also referred to as the same as position picture (colPic).Movement Information may include motion vector and reference picture indices.The information of such as prediction mode information and motion information can be compiled by (entropy) Code, then output is the form of bit stream.

When using the motion information of time contiguous block under skip mode and merging patterns, in reference pictures list most High picture can be used as reference pictures.It can be drawn based on current picture with corresponding benchmark including the reference pictures in reference pictures list Picture order count (POC) between face is poor and is aligned.POC corresponds to display order and can distinguish with coding orders.

Subtracter 115 generates the residual sample of the difference as original sample and forecast sample.It, can when application skip mode Residual sample is not generated as described above.

Converter 120 converts residual sample as unit of transform block, to generate transformation coefficient.Converter 120 can be based on The size of correspondent transform block and prediction mode applied to the encoding block or prediction block that spatially overlap with transform block are held Row transformation.For example, when being 4 × 4 residual errors to the encoding block or prediction block application intra prediction and transform block that are overlapped with transform block Discrete sine transform (DST) Lai Bianhuan can be used in residual sample when array, uses discrete cosine transform (DCT) in other cases To convert.

Quantization of transform coefficients can be generated the transformation coefficient of quantization by quantizer 125.

The transformation coefficient of 130 pairs of rearrangement device quantizations is reset.Rearrangement device 130 can be by coefficient scanning method by block form The transformation coefficient of quantization be rearranged to a n dimensional vector n.Although rearrangement device 130 is described as individual component, rearrangement device 130 It can be a part of quantizer 125.

Entropy coder 135 can execute entropy coding to the transformation coefficient of quantization.Entropy coding may include such as Exp-Golomb (Golomb), context-adaptive variable length code (CAVLC), context adaptive binary arithmetic coding (CABAC) etc. Coding method.Other than the transformation coefficient of quantization, entropy coder 135 can be to the information needed for video reconstruction (for example, syntax Element value etc.) together or it is executed separately coding.The information of entropy coding can be with network abstract layer (NAL) according to bit stream form Unit is sent or storage.

The value (transformation coefficient) that inverse quantizer 140 quantifies quantizer 125 carries out inverse quantization, and inverse converter 145 is to inverse The inverse-quantized value of quantizer 135 carries out inverse transformation to generate residual sample.

Residual sample is added with forecast sample with reconstruction picture by adder 150.Residual sample can in blocks and in advance This addition of test sample is to generate reconstructed blocks.Although adder 150 is described as individual component, adder 150 can be pre- Survey a part of device 110.

Filter 155 can adaptively deviate the picture application de-blocking filter and/or sample of reconstruct.In the picture of reconstruct The distortion in artifact or quantization at block boundary can be adaptively deviated by de-blocking filter and/or sample to correct.It can deblock Filtering is adaptively deviated as unit of sample using sample after completing.Filter 155 can be adaptive to the application of the picture of reconstruct Loop filter (ALF).It can be to the picture application ALF for the reconstruct for having applied de-blocking filter and/or sample adaptively to deviate.

Information needed for memory 160 can store the picture or coding/decoding of reconstruct.Here, the picture of reconstruct can be By the picture for the reconstruct that filter 155 filters.The picture of the reconstruct stored can be used as predicting for (interframe) of other pictures Reference pictures.For example, memory 160 can store (benchmark) picture for inter-prediction.It here, can be according to reference pictures collection It closes or the picture for inter-prediction is specified in reference pictures list.

Fig. 2 schematically illustrates the structure for the video decoder that the disclosure is suitable for.

Referring to Fig. 2, video decoder 200 includes entropy decoder 210, rearrangement device 220, inverse quantizer 230, inverse converter 240, fallout predictor 250, adder 260, filter 270 and memory 280.

When input include video information bit stream when, video decoder 200 can in video coding apparatus to video The procedure correlation that information is handled carrys out reconstructing video.

For example, the processing unit applied in video coding apparatus can be used to execute video decoding in video decoder 200. Therefore, the decoded processing unit block of video can be coding unit block, predicting unit block or converter unit block.As decoded list First block, coding unit block can be divided according to quad-tree structure from maximum coding unit block.As what is divided from coding unit block Block, predicting unit block can be the cell block of sample predictions.In this case, predicting unit block is divided into sub-block.Make For coding unit block, converter unit block can be divided according to quad-tree structure, and can be the list for deriving transformation coefficient First block or for from transformation coefficient derive residual signals cell block.

Entropy decoder 210 can parse bit stream to export video reconstruction or picture and reconstruct required information.For example, entropy decoding Device 210 can the coding method based on exp-Golomb coding, CAVLC, CABAC etc. the information in bit stream is solved Code, and the quantized value of the value of syntax elements needed for exportable video reconstruction and the transformation coefficient about residual error.

More specifically, CABAC entropy decoding method can receive bin corresponding with each syntax elements in bit stream, use Decode target syntax elements information and it is neighbouring and decoding object block decoded information or in previous steps decoded symbol/ The information of bin determines context model, bin generating probability is predicted according to identified context model, and execute bin's Arithmetic decoding is to generate symbol corresponding with each syntax elements value.Here, CABAC entropy decoding method can determine context mould Context model is updated using the information for the decoded symbol/bin of following symbol/bin context model after type.

Information in entropy decoder 210 in decoded information about prediction is provided to fallout predictor 250, and entropy solution The residual values (that is, transformation coefficient of quantization) that code device 210 has executed entropy decoding can be input to rearrangement device 220.

The transformation coefficient of quantization can be rearranged to two-dimensional block form by rearrangement device 220.Rearrangement device 220 can be performed and code device The corresponding rearrangement of performed coefficient scanning.Although rearrangement device 220 is described as individual component, rearrangement device 220 can be with It is a part of quantizer 230.

Inverse quantizer 230 can carry out inverse quantization based on transformation coefficient of (inverse) quantization parameter to quantization with output transform system Number.In this case, the information for deriving quantization parameter can be signaled from code device.

Inverse converter 240 can carry out inverse transformation to transformation coefficient to derive residual sample.

Fallout predictor 250 can execute prediction to current block, and produce the prediction block of the forecast sample including current block.In advance The unit for surveying the prediction executed in device 250 can be encoding block or can be transform block or can be prediction block.

Fallout predictor 250 can determine whether to apply intra prediction or inter-prediction based on the information about prediction.This In the case of, which unit will be used to may differ from being used to generate pre- test sample between intra prediction and inter-prediction for determining This unit.In addition, the unit for generating forecast sample can also be different in inter-prediction and intra prediction.For example, interframe Which will can be determined as unit of CU using between prediction and intra prediction.In addition, for example, can lead in inter-prediction It crosses as unit of PU and determines prediction mode to generate forecast sample, in intra prediction, can be predicted by being determined as unit of PU Mode generates forecast sample as unit of TU.

In the case where intra prediction, fallout predictor 250 can be derived currently based on the neighbouring baseline sample in current picture The forecast sample of block.Fallout predictor 250 can pass through neighbouring baseline sample application directional pattern or non-directional mode based on current block To derive the forecast sample of current block.In this case, the intra prediction mode of contiguous block can be used to be applied to determine The prediction mode of current block.

In the case where inter-prediction, fallout predictor 250 can be based on the sample specified in reference pictures according to motion vector To derive the forecast sample of current block.One of skip mode, merging patterns and MVP mode can be used to work as to derive for fallout predictor 250 Preceding piece of forecast sample.Here, it can obtain or derive based on the information about prediction and is current provided by video coding apparatus Motion information needed for the inter-prediction of block (for example, motion vector and information about reference picture indices).

Under skip mode and merging patterns, the motion information of contiguous block can be used as the motion information of current block.Here, adjacent Nearly block may include spatial neighbor block and time contiguous block.

Fallout predictor 250, which can be used, to be arranged with the motion information of contiguous block come juxtaposition candidate list and using candidate is merged Merge motion vector of the indicated information of index as current block on table.Merging index can be signaled by code device. Motion information may include motion vector and reference pictures.When the movement for using time contiguous block under skip mode and merging patterns When information, the highest picture in reference pictures list can be used as reference pictures.

In the case where skip mode, is mutually distinguished with merging patterns, do not send the difference between forecast sample and original sample (residual error).

In the case where MVP mode, it is current to derive as motion vector predictor that the motion vector of contiguous block can be used The motion vector of block.Here, contiguous block may include spatial neighbor block and time contiguous block.

When application merging patterns when, for example, can be used reconstruct spatial neighbor block motion vector and/or with as the time The corresponding motion vector of Col block of contiguous block generates merging candidate list.Under merging patterns, selected from merging candidate list The motion vector of candidate blocks is used as the motion vector of current block.The above-mentioned information about prediction may include indicating to be selected to be included in conjunction And the merging index of the candidate blocks with optimum movement vector of the candidate blocks in candidate list.Here, fallout predictor 250 can be used Merge index to derive the motion vector of current block.

When as another sample application MVP (motion-vector prediction) mode, the fortune of the spatial neighbor block of reconstruct can be used Dynamic vector and/or motion vector corresponding with the Col block as time contiguous block generate motion vector predictor candidate list. That is, the motion vector and/or motion vector corresponding with the Col block as time contiguous block of the spatial neighbor block of reconstruct can be used as Motion vector candidates.The above-mentioned information about prediction may include instruction selected from including the best of motion vector candidates in lists The predicted motion vector index of motion vector.Here, motion vector index can be used to wait from motion vector is included in for fallout predictor 250 The motion vector candidates in list are selected to select the predicted motion vector of current block.The fallout predictor of code device can get current block Motion vector difference (MVD) between motion vector and motion vector predictor, encodes MVD, and in the form of bit stream The MVD of exports coding.That is, MVD can be obtained by subtracting motion vector predictor from the motion vector of current block.Here, in advance Surveying device 250 can obtain including the motion vector in the information about prediction and by by motion vector difference and motion-vector prediction The motion vector of the Calais Qi Xiang derivation current block.In addition, fallout predictor can be from the above-mentioned information acquisition about prediction or derivation instruction The reference picture indices of reference pictures.

Residual sample can be added with forecast sample to reconstruct current block or current picture by adder 260.Adder 260 can By the way that residual sample and forecast sample phase Calais are reconstructed current picture in blocks.When application skip mode, do not send Residual error, therefore forecast sample can be changed to reconstructed sample.Although adder 260 is described as individual component, adder 260 It can be a part of fallout predictor 250.

Filter 270 can to the picture application de-blocking filter of reconstruct, sample adaptively deviates and/or ALF.It here, can be It is adaptively deviated as unit of sample using sample after de-blocking filter.It can be adaptively inclined in de-blocking filter and/or application sample ALF is applied after moving.

Memory 280 can store the picture of reconstruct or the information that decoding is required.Here, the picture of reconstruct can be by filtering The picture for the reconstruct that device 270 filters.For example, memory 280 can store the picture for inter-prediction.Here, pre- for interframe The picture of survey can be specified according to reference pictures set or reference pictures list.The picture of reconstruct can be used as the benchmark of other pictures Picture.Memory 280 can export the picture of reconstruct according to output order.

As described above, the motion information of object block can pass through application in the case where executing inter-prediction for object block Skip mode, merging patterns or adaptive motion vector predict (AMVP) mode to generate and be encoded and export.In this feelings Under condition, in the motion information of object block, due to the process encoded in blocks, can calculate and including distortion, it is thereby possible to Without the motion information of the perfect reconstructed blocks for reflecting expression object block.Specifically, to object block application merging patterns the case where Under, the accuracy of the motion information of object block may deteriorate.That is, in the prediction block and mesh of the motion information derivation by object block There are greatest differences between the reconstructed blocks of mark block.In this case, in next piece of the decoding process neighbouring with object block Using the motion information of object block, and it is distorted possible propagation, overall coding efficiency can deteriorate as a result,.

Therefore, in the present invention, a kind of method is proposed, wherein based on the weight of derivation after the decoding process of object block Building block calculates the modified motion information of object block, and the motion information of object block is updated based on modified motion information, is made More accurate motion information can be derived by obtaining next piece neighbouring with object block.Overall coding efficiency can improve as a result,.

Fig. 3 is shown based on one-way movement information come the case where executing inter-prediction and based on applied to the two-way of object block Motion information is come example the case where executing inter-prediction.Bi-directional motion information may include L0 reference picture indices and L0 movement arrow Amount and L1 reference picture indices and L1 motion vector, one-way movement information may include L0 reference picture indices and L0 motion vector Or L1 reference picture indices and L1 motion vector.L0 indicates reference pictures list L0 (list 0), and L1 indicates reference pictures list L1 (list 1).In image encoding process, the method for inter-prediction may include being derived by Motion estimation and compensation Motion information.As shown in figure 3, estimation can be for the benchmark encoded before the coding of the target picture including object block Picture indicates during deriving with matched piece of object block.It can be defined as reference block with matched piece of object block, pass through vacation If in target picture including reference block identical with object block, the alternate position spike between the object block derived and reference block can be determined Justice is the motion vector of object block.Motion compensation may include deriving and using the one-way method of reference block and derivation and making With the two way method of two reference blocks.

The information of the information of motion vector and reference pictures including object block can be defined as the motion information of object block. The information of reference pictures may include reference pictures list and instruction includes the benchmark picture of the reference pictures in reference pictures list Face index.Encoding device can for after object block is encoded will then target block coding next piece neighbouring with object block Or will then target picture coding picture storage object block motion information.The motion information stored can indicate to connect Next piece of target block coding or then to use in the method for the motion information of the picture of target block coding.This method can wrap Include: merging patterns are indexed the motion information of object block and send next piece of the motion information neighbouring with object block and make For the method for index；And AMVP mode, it is indicated merely with the difference of next piece of motion vector and the motion vector of object block The method of next piece of the motion vector neighbouring with object block.

The invention proposes a kind of to update target picture or object block after the cataloged procedure of target picture or object block Motion information method.When executing inter-prediction for object block and encoding, it is storable in fortune used in inter-prediction Dynamic information.However, motion information may include the distortion occurred during the process calculated by block matching method, and due to fortune Dynamic information can be the value for distortion (RD) the optimization process selection of object block percent of pass, it is possible that being difficult to ideally reflect mesh Mark the actual motion of block.Since the motion information of object block can be used not only in the inter predication process of object block, and influence It is including encoding after the coding of next piece of the target picture neighbouring with object block encoded after the coding of object block Picture and target picture including next piece He object block, so overall coding efficiency can deteriorate.

Fig. 4 shows the example of the cataloged procedure of the method for the motion information including updating object block.Encoding device is to target Block is encoded (step S400).Encoding device can derive forecast sample and base by executing inter-prediction for object block The reconstructed blocks of object block are generated in forecast sample.Encoding device can derive the movement letter of the object block for executing inter-prediction Cease and generate the information of the inter-prediction of the object block including the motion information.Motion information is referred to alternatively as the first movement letter Breath.

After the cataloged procedure of performance objective block, encoding device calculates the modified movement letter of the reconstructed blocks based on object block It ceases (step S410).Encoding device can calculate the modified motion information of object block by various methods.Modified movement letter Breath is referred to alternatively as the second motion information.Direct method including light stream (OF) method, block matching method, frequency domain method etc. with And singular point matching process, using at least one method of the indirect method of the method for statistical property etc. it can be applied to the party Method.In addition, direct method and indirect method can be applied simultaneously.The detailed content of block matching method and OF method is described below.

Due to the characteristic that the motion information of object block calculates, encoding device is likely difficult to the letter of the coding merely with object block Breath is to calculate more accurate motion information, thus, for example, the time that encoding device calculates modified motion information can execute packet After the cataloged procedure for including the target picture of object block, without after the cataloged procedure of object block.That is, encoding device is executable It the cataloged procedure of target picture and is calculated compared with after the cataloged procedure of performance objective block based on more information Modified motion information.

In addition, encoding device also produce instruction include existing motion vector in (first) motion information with it is modified The motion vector of difference between motion information updates poor information and is encoded and exported to it.Motion vector updates poor information can It is sent as unit of PU.

Encoding device determines whether to update the motion information (step S420) of object block.Encoding device can be transported by (first) Dynamic information determines whether to update (first) motion information compared with the accuracy between modified motion information.For example, compiling Decoding apparatus can be used the amount of the difference between the image and original image that derive by motion compensation using each motion information come really It is fixed whether to update (first) motion information.In other words, encoding device can be by comparing the benchmark derived based on each motion information The data volume of residual signals between block and the original block of object block determines whether to update.

In the step s 420, in the case where determining the motion information for updating object block, encoding device is based on modified fortune Dynamic information updates the motion information of object block and stores modified motion information (step S430).For example, based on modified The data volume for the residual signals between special datum block and original block that motion information derives, which is less than, is based on (first) motion information In the case where the data volume of residual signals between the reference block and original block of derivation, encoding device can be believed based on modified movement Breath is to update (first) motion information.In this case, encoding device can be by replacing with amendment for (first) motion information Motion information come update object block motion information and store only including modified motion information update motion information.Separately Outside, encoding device can be by updating the motion information of object block simultaneously for modified motion information and (first) motion information phase Calais Storage includes the motion information of the update of (first) motion information and modified motion information.

In addition, in the step s 420, in the case where determination does not update the motion information of object block, encoding device storage (the first) motion information (step S440).For example, the special datum block that is derived based on modified motion information and original block it Between residual signals data volume be not less than based on (first) motion information derive reference block and original block between residual error letter Number data volume in the case where, encoding device can store (first) motion information.

On the other hand, although being not shown, modified motion information is being derived without by used in inter-prediction (the One) in the case that motion information determines update compared between modified motion information, encoding device can be based on modified fortune Dynamic information updates the motion information of object block.

Further, since decoding device is not useable for using original image, indicate whether to update so decoding device can receive The additional information of object block.That is, encoding device produces and encodes the additional information for indicating whether to update, and defeated by bit stream Out it.For example, indicating whether that the additional information updated is referred to alternatively as updating mark.Movement letter can be indicated by updating the case where block is 1 Breath is updated, and updating the case where block is 0 can indicate that motion information does not update.For example, updating mark can be sent as unit of PU.Separately Selection of land, updating mark can send as unit of CU, as unit of CTU or as unit of slice, and can be by such as drawing The higher level of the unit of the unit or sequence parameter set (SPS) of face parameter set (PPS) is sent.

In addition, (first) motion information and modified fortune that decoding device can pass through object block based on the reconstructed blocks of object block The comparison between information is moved to determine whether to update, updates mark without receiving, and in the motion information quilt for determining object block In the case where update, decoding device can be updated the motion information of object block based on modified motion information and store the fortune of update Dynamic information.In addition, deriving modified motion information without by motion information used in inter-prediction and modified movement In the case that comparison between information determines update, decoding device can update the movement of object block based on modified motion information Information.In this case, decoding device can update mesh by the way that (first) motion information is replaced with modified motion information It marks the motion information of block and stores the motion information of the only update including modified motion information.In addition, decoding device can pass through It includes that (first) is transported that modified motion information is updated the motion information of object block and stored with (first) motion information phase Calais The motion information of the update of dynamic information and modified motion information.

In the case where executing the process for updating motion information after the cataloged procedure of object block, can be performed under object block Next piece of the cataloged procedure neighbouring with object block of one coding orders.

Fig. 5 shows the example of the decoding process of the method for the motion information including updating object block.Decoding process can be according to Similar with above-mentioned cataloged procedure mode executes.Decoding device is decoded (step S500) to object block.To object block In the case where inter-prediction, decoding device can obtain the information of the inter-prediction for object block by bit stream.Decoding Equipment can be come the motion information of derived object block and pre- by executing interframe for object block based on the information for inter-prediction It surveys to derive forecast sample.Motion information is referred to alternatively as the first motion information.Decoding device can generate mesh based on forecast sample Mark the reconstructed blocks of block.

The modified motion information (step S510) of decoding device calculating object block.Decoding device can come by various methods Calculate modified motion information.Modified motion information is referred to alternatively as the second motion information.Including such as light stream (OF) method, block The direct method and singular point matching process of matching process, frequency domain method etc. use the indirect of the method for statistical property etc. At least one method of method can be applied to this method.In addition, direct method and indirect method can be applied simultaneously.It is described below The detailed content of block matching method and OF method.

Due to the characteristic that the motion information of object block calculates, decoding device is likely difficult to the decoded letter merely with object block Breath is to calculate more accurate motion information, thus, for example, the time that decoding device calculates modified motion information can execute packet After the decoding process for including the target picture of object block, without after the decoding process of object block.That is, decoding device is executable It the decoding process of target picture and is calculated compared with after the decoding process of performance objective block based on more information Modified motion information.

In addition, it includes existing in (first) motion information that decoding device can also obtain generation instruction by bit stream The motion vector of difference between motion vector and modified motion information updates poor information.Motion vector updates poor information can be with PU For unit transmission.In this case, decoding device can not independently calculate amendment by above-mentioned direct method and indirect method Motion information, but derive amendment by updating poor information summation to (first) motion information and motion vector obtained Motion information.That is, decoding device can derive existing movement using the motion vector predictor (MVP) of modified motion information Vector simultaneously derives modified motion vector with existing motion vector phase Calais by the way that motion vector is updated poor information.

Decoding device determines whether to update the motion information (step S520) of object block.Decoding device can be transported by (first) Dynamic information determines whether to update (first) motion information compared with the accuracy between modified motion information.For example, solution Decoding apparatus, which can be used, uses the difference between each motion information reference block derived and the reconstructed blocks of object block by motion compensation Amount come determine whether update (first) motion information.In other words, encoding device can be pushed away by comparing based on each motion information The data volume of residual signals between the original block of the reference block and object block led determines whether to update.

In addition, decoding device is received the additional information for indicating whether to update from encoding device and is determined based on additional information Whether the motion information of object block is updated.For example, indicating whether that the additional information updated is referred to alternatively as updating mark.Updating block is 1 the case where, can indicate that motion information is updated, and updating the case where block is 0 can indicate that motion information does not update.For example, updating mark It can be sent as unit of PU.

In step S520, in the case where determining the motion information for updating object block, decoding device is based on modified fortune Dynamic information updates the motion information of object block and stores the motion information (step S530) of update.For example, based on modified The data volume of residual signals between the special datum block that motion information derives and reconstructed blocks, which is less than, is based on (first) motion information In the case where the data volume of residual signals between the reference block and reconstructed blocks of derivation, decoding device can be believed based on modified movement Breath is to update (first) motion information.In this case, decoding device can be by replacing with amendment for (first) motion information Motion information come update object block motion information and store only including modified motion information update motion information.Separately Outside, decoding device can be by updating the motion information of object block simultaneously for modified motion information and (first) motion information phase Calais Storage includes the motion information of the update of (first) motion information and modified motion information.

In addition, in step S520, in the case where determination does not update the motion information of object block, decoding device storage (the first) motion information (step S540).For example, the special datum block that is derived based on modified motion information and reconstructed blocks it Between residual signals data volume be not less than based on (first) motion information derive reference block and reconstructed blocks between residual error believe Number data volume in the case where, decoding device can store (first) motion information.

On the other hand, although being not shown, modified motion information is being derived without by (first) motion information and amendment Motion information between comparison determine update in the case where, decoding device can update object block based on modified motion information Motion information.

In addition, the motion information stored and the motion information being used for transmission can have different resolution ratio.In other words, it repairs The unit of included motion vector and based in the motion information for the information inference of inter-prediction in positive motion information The unit of included motion vector can be different.For example, the movement of the motion information based on the information inference for inter-prediction The unit of vector can have the unit of 1/4 part sample, the motion vector of the modified motion information calculated in decoding device Unit can have the unit of 1/8 part sample or 1/16 part sample.Decoding device can be at the place for calculating modified motion information It is to store required resolution ratio, or pass through calculating in the processing for storing modified motion information for resolution adjustment in reason (for example, round up, multiplication etc.) adjusts resolution ratio.Modified motion information motion vector high resolution in being based on In the case where the motion vector of the motion information of information inference for inter-prediction, for by the time adjacent to motion information based on The zoom operations of calculation, accuracy can increase.In addition, different with the resolution ratio for inside operation being used for transmission motion information In the case where decoding device, decoded effect can be executed according to internal operation standard by existing.

In addition, in the case where applying block matching method in the method for calculating modified motion information, it is modified Motion information can derive as follows.Block matching method may be expressed as method for estimating used in encoding device.

Encoding device can be by measuring distortion journey according to the difference between object block and the accumulation of phase sample of reference block It spends and these is used as cost function, then, derive and the most like reference block of object block.That is, encoding device can be based on and mesh The smallest reference block of residual error for marking the reconstructed blocks (or original block) of block carrys out the modified motion information of derived object block.Residual error is minimum Reference block be referred to alternatively as special datum block.Absolute difference and (SAD) and mean square error (MSE) can be used for indicating the difference between sample The function of value.By measuring distortion level according to the absolute value of the difference between object block and the accumulation of phase sample of reference block SAD can be derived based on following formula.

[formula 1]

Herein, Block_cur(i, j) indicates the reconstruct sample of (i, j) coordinate in the reconstructed blocks (or original block) of object block This (or original sample), Block_ref(i, j) indicate reference block in (i, j) coordinate reconstructed sample, width be reconstructed blocks (or Original block) width, height indicate reconstructed blocks (or original block) height.

In addition, by measuring distortion according to the square value of the difference between object block and the accumulation of phase sample of reference block The MSE of degree can be derived based on following formula.

[formula 2]

Computation complexity for calculating the method for modified motion information can be according to for searching for searching for special datum block Rope range neatly changes.Therefore, in the case where calculating the modified motion information of object block using block matching method, solution Decoding apparatus can search only for including away from the benchmark derived by (first) motion information used in the decoding process in object block Block is the reference block in presumptive area, can maintain low computation complexity as a result,.

Fig. 6 shows the example for updating the method for motion information of object block by block matching method by decoding device.Decoding Equipment is decoded (step S600) to object block.To object block application inter-prediction, decoding device can pass through Bit stream obtains the information of the inter-prediction for object block.Due to being applied to the movement letter of the update object block of encoding device The method of breath should be applied to decoding device in the same fashion, so encoding device and decoding device can be based on object blocks Reconstructed blocks calculate the modified motion information of object block.

Decoding device executes estimation (step for modified motion information using reconstructed blocks in reference pictures S610).Decoding device can derive the reference pictures of the motion information based on the information inference for inter-prediction as modified The particular reference picture of motion information, and particular reference picture is detected in the reference block in reference pictures.Special datum block The summation (that is, absolute difference and (SAD)) that can be the absolute value of the difference in reference block between the sample of reconstructed blocks is minimum Reference block.In addition, the region of search for being used to detect special datum block can be limited to away from by being based on being used for frame by decoding device Between the presumptive area of reference block that indicates of the motion information of information inference predicted.In other words, decoding device can derive be located at away from Reference block is the reference block in presumptive area in the middle with the smallest reference block of SAD of reconstructed blocks as special datum block.Decoding is set Standby can only be to execute estimation in presumptive area away from reference block, increase amendment while reducing computation complexity as a result, Motion information reliability.

In the case where the size of the reconstructed blocks of object block is greater than particular size, in order to carefully calculate modified movement letter Reconstructed blocks are divided into size to be less than the fritter of particular size and calculate more detailed movement based on each fritter by breath, decoding device Information (step S620).Particular size can be pre-configured with, and the block divided is referred to alternatively as sub- reconstructed blocks.Decoding device can be by weight Building block is divided into multiple sub- reconstructed blocks and is reconstructed based on sub- reconstructed blocks in the reference pictures of modified motion information with each height Block is that unit derives specific sub- reference block, and the specific sub- reference block derived is as special datum block.Decoding device can Modified motion information is calculated based on special datum block.

In addition, in the case where applying light stream (OF) method in the method for calculating modified motion information, amendment Motion information can derive as follows.OF method can be calculated based on such hypothesis: the speed of the object in object block is uniformly simultaneously And indicate that the sample value of the sample of the object does not change in the picture.δ x is moved in x-axis in time δ t in object and in y-axis In the case where upper mobile δ y, following formula can be established.

[formula 3]

I (x, y, t)==I (x+ δ x, y+ δ y, t+ δ t)

Herein, I (x, y, t) instruction expression includes the reconstruct in position (x, y) of the object of the time t in object block The sample value of sample.

In formula 3, when right item is unfolded in Taylor series, following formula can be derived.

[formula 4]

When meeting formula 4, following formula can be established.

[formula 5]

Rewriting formula 5 can derive following formula.

[formula 6]

Herein, v_xIt is vector component of the motion vector calculated in x-axis, v_yIt is motion vector calculated in y-axis On vector component.Decoding device can derive local derviation value of the object on x-axis, y-axis and t axis, and the local derviation by will be derived Value is applied to the formula to derive the motion vector (v of current location (that is, position of reconstructed sample)_x,v_y).In this case, example Such as, decoding device can configure the reconstructed sample in the reconstructed blocks for indicating the object block of object to include in the region of 3 × 3 sizes Reconstructed sample in unit, and left item is calculated close to 0 motion vector (v by the way that the formula is applied to reconstructed sample_x,v_y)。

The storage format of the modified motion information of the object block calculated by encoding device can have various formats, due to it It can be used for the picture encoded after the cataloged procedure for the target picture for including in next piece neighbouring with object block or object block, It is possible that it is beneficial that modified motion information can be stored according to format identical with the motion information for predicting object block. Motion information for prediction may include about the information, reference picture indices and motion vector for being single prediction or double prediction. Modified motion information can be calculated as having format identical with the motion information for prediction.

There is only target picture a reference pictures, that is, in addition to the picture order count of target picture (POC) except the case where value is 1, double prediction is executable.In general, being compiled compared with the case where encoding device executes single prediction Estimated performance can be higher in the case that decoding apparatus executes double prediction.Therefore, it can be used for the case where executing double prediction in target picture Under, in order to improve the overall coding efficiency of image by propagating accurate motion information, encoding device can pass through above method benefit The modified motion information of object block is calculated and updated with double prediction.However, even if can be used for executing in target picture double pre- In the case where survey, that is, even if can also block in the case where the value of the POC of target picture is not 1 and can occur can not The case where executing double prediction.Believed when based on modified motion information compensation campaign and by motion compensation based on modified movement The sum of the sample of the forecast sample derived and the absolute value of the difference of sample of reconstructed blocks (or original block) of object block is ceased to be greater than in advance When determining threshold value, it may be determined that block.For example, the case where the modified motion information of object block is double prediction motion information Under, when the reconstruct based on the sample and object block that include the forecast sample that the L0 motion vector in modified motion information derives When the sum of absolute value of the difference of sample of block (or original block) is greater than predetermined threshold, according to the prediction to L0, it may be determined that hide Gear.In addition, when based on the sample and object block for including the forecast sample that the L1 motion vector in modified motion information derives The sum of the absolute value of the difference of sample of reconstructed blocks (or original block) when being greater than predetermined threshold, according to the prediction to L1, it may be determined that It blocks.In the case where being blocked by the prediction to any of L0 and L1, in addition to the list blocked Except the information of prediction, encoding device can derive modified motion information as single prediction motion information.

Modified motion information is being blocked or can be used for calculating by the prediction to any of L0 and L1 In the case of, the method that the motion information of the contiguous block of object block can be used to apply the motion information for updating object block, alternatively, The method for not updating the motion information of object block can be applied based on modified motion information.In order to store and propagate accurate fortune Dynamic information, in these cases, the method for not updating the motion information of object block based on modified motion information may be more appropriate.

The L0 reference picture indices and L1 reference picture indices of modified motion information can indicate in reference pictures list One of reference pictures (L0 or L1).Reference pictures indicated by the reference picture indices of modified motion information are referred to alternatively as specific Reference pictures.Select the method for one of several reference pictures in reference pictures list can be as described below.

For example, encoding device may be selected to include that the benchmark encoded recently in the reference pictures in reference pictures list is drawn Face.That is, it includes the reference picture indices in modified motion information that encoding device, which produces, instruction is included in reference pictures The reference pictures encoded recently in reference pictures in list.

In addition, the picture order count for having closest to current picture in reference pictures list may be selected in encoding device (POC) reference pictures of POC.That is, it includes the reference picture indices in modified motion information that encoding device, which produces, Instruction include in the reference pictures in reference pictures list POC with include object block target picture POC difference it is exhausted To the smallest reference pictures of value.

In addition, the reference pictures of the lower level belonged in layered structure in reference pictures list may be selected in encoding device.Belong to It can be I slice in the reference pictures of lower level or by the reference pictures of application low quantization parameter (QP) coding.

In addition, the highest reference pictures of reliability of motion compensation in reference pictures list may be selected in encoding device.That is, compiling Decoding apparatus can in the reference block being located in reference pictures included in reference pictures list derived object block reconstructed blocks The special datum block of (or original block) and to generate include the reference picture indices in modified motion information, instruction includes institute The reference pictures of the special datum block of derivation.

The method of the reference picture indices of the above-mentioned modified motion information of generation can independently apply or these methods can Combined application.

It can be by including OF method, block matching method, frequency domain method including the motion vector in modified motion information Deng at least one method derive, and may need with the motion vector as unit of at least smallest blocks.

Referring to Fig. 7, in the case where target picture is the picture that POC value is 3, double prediction can be executed for target picture, The picture that the picture and POC value that POC value is 2 are 4 can be derived the reference pictures as target picture.The L0 of target picture can be wrapped The picture that POC value is 2 and the picture that POC value is 0 are included, the corresponding respective value of L0 reference picture indices can be 0 and 1.In addition, target The picture that the L1 of picture may include the picture that POC value is 4 and POC value is 8, the corresponding respective value of L1 reference picture indices can be 0 With 1.In this case, the particular reference picture of the modified motion information of target picture can be selected as each by encoding device The smallest reference pictures of the absolute value of the difference of POC and target picture in reference pictures in reference pictures list.In addition, coding Equipment can derive the motion vector of modified motion information in target picture using the block of 4 × 4 sizes as unit.OF method can The method for being applied to derive motion vector.Motion vector based on POC value for 2 reference pictures derivation can be stored in amendment Motion information in included L0 motion information, the motion vector that derives of reference pictures for being 4 based on POC value can be stored In modified motion information in included L1 motion information.

In addition, in the case where target picture is the picture that POC value is 8, single prediction can be performed in target picture referring to Fig. 7, And the picture that POC value is 0 can be derived the reference pictures as target picture.The L0 of target picture may include that POC value is 0 Reference pictures, and the value of L0 reference picture indices corresponding with reference pictures can be 0.In this case, encoding device can The particular reference picture of the modified motion information of target picture is selected as including POC and mesh in the reference pictures in L0 Mark the smallest reference pictures of absolute value of the difference of the POC of picture.In addition, encoding device can be in current picture with 4 × 4 sizes Block is the motion vector that unit derives modified motion information.OF method can be applied to the method for deriving motion vector.Block Method of completing the square can be applied to the method for deriving motion vector, and the motion vector based on POC value for 0 reference pictures derivation can be deposited It stores up in included L0 motion information in modified motion information.

Referring to above embodiment, the reference pictures of the modified motion information of target picture can be derived as instruction and include The smallest reference pictures of the absolute value of the difference of POC and the POC of target picture in the reference pictures in reference pictures list, and And the motion vector of modified motion information can be calculated using the block of 4 × 4 sizes as unit.It is the picture that POC is 8 in target picture In the case where face, since target picture is the picture that inter-prediction is first carried out in picture shown in Fig. 7, so executing single pre- It surveys, but in the case where target picture is the picture that POC is 3, double prediction can be used and the modified movement of target picture letter Breath can be confirmed as double prediction motion information.In addition, even if being drawn in the case where target picture is the picture that POC is 3 in target In the case where calculating in blocks in face and updating modified motion information, encoding device can determine selection packet in blocks Include a motion information format of double prediction motion information and single prediction motion information.

In addition, being different from above embodiment, the reference picture indices of modified motion information are produced to indicate by dividing The reference pictures of layer structure choice, rather than the POC including POC and target picture in the reference pictures in reference pictures list The smallest reference pictures of absolute value of the difference.For example, target picture is repaired in the case where target picture is the picture that POC is 5 The L0 reference picture indices of positive motion information can indicate that POC value is 0 picture, rather than the picture that POC value is 4.Executing mesh The timing after the cataloged procedure of picture is marked, the major part of the reference pictures encoded recently can indicate the POC of POC and target picture The smallest reference pictures of absolute value of the difference.However, the case where being similar to the picture that POC value is 6, the reference pictures encoded recently It is the reference pictures that POC value is 2, and it is 4 that the smallest picture of the absolute value of the difference of POC and the POC of target picture, which is POC value, Picture therefore can there is a situation where indicate different reference pictures.

In the case where the motion information of object block is updated to modified motion information by the above method, it can be used and repair To execute next piece of the inter-prediction neighbouring with object block, (object block is referred to alternatively as first piece, next piece to positive motion information It is referred to alternatively as second piece).It the use of the method for modified motion information may include to amendment under next piece of inter-frame forecast mode Motion information be indexed and the method that sends and by next piece of motion vector be expressed as the motion vector of object block with The method of difference between next piece of motion vector.Modified motion information is indexed and is sent by a method of frame Between in prediction mode to the method in the case where next piece of application merging patterns, next piece of motion vector is expressed as target The method of difference between the motion vector of block and next piece of motion vector can be in inter-frame forecast mode to next piece Using the method in the case where AMVP mode.

Fig. 8 show in the case where the motion information of object block is updated to the modified motion information of object block with target The example of next piece neighbouring of merging candidate list of block.Referring to Fig. 8, configurable next piece including object block of encoding device Merge candidate list.The transmittable instruction of encoding device includes merging the motion information in the block in candidate list with next piece The merging of most like block indexes.In this case, the motion information of the update of object block can be used for next piece of space neighbour The motion information of nearly candidate blocks, or alternatively, the motion information for time neighbor candidate block.The motion information of object block can Modified motion information including object block.As shown in figure 8, the motion information of spatial neighbor candidate blocks can indicate A1, B1, B0, On the position A0 and B2 with a motion information in the motion information of next piece of neighbouring block stored.It may be updated next The motion information of the object block encoded before the coding of block, and the motion information updated can under the influence of one piece of cataloged procedure. In the case where the motion information for including mistake travels to next piece by merging patterns, which may accumulate and propagate, but It is that can reduce the mistake of propagation by updating the motion information in each piece via the above method.

Fig. 9 show in the case where the motion information of object block is updated to the modified motion information of object block with target The example of next piece neighbouring of merging candidate list of block.It can store for predicting and (first) of next piece of neighbouring object block Motion information and modified motion information.In this case, encoding device can will indicate in next piece of merging candidate list The object block of the modified motion information of the object block newly calculated is configured to individually merge candidate blocks.As shown in figure 9, instruction mesh The object block for marking the modified motion information of block can be inserted into merging candidate list, instruction and next piece of neighbouring object block Beside the object block of (first) motion information applied in cataloged procedure.In addition, the changeable priority for merging candidate list, and And the quantity of insertion position can be configured differently.

Although the example as the object block for deriving modified motion information shows A1 and B1, this is only example, Even for remaining A0, B0, B2, T0 and T1, modified motion information can also be derived, and it is candidate that merging can be configured based on it List.

For example, next piece can be located on the picture different from object block, and next piece of time neighbor candidate block can wrap It includes in merging candidate list.The motion information of time neighbor candidate block can be selected by next piece of merging index.At this In the case of kind, the motion information of time neighbor candidate block can be updated to modified motion information by the above method, and refer to Show that the time neighbor candidate block of the motion information of update can be inserted in next piece of merging candidate list.In addition, by upper Method is stated, (first) motion information and modified motion information of time neighbor candidate block can be stored separately.In such case Under, during generating next piece of merging candidate list, indicate that the time neighbor candidate block of modified motion information can be made For the additional candidate insertion for merging candidate list.

Different from content shown in Fig. 9, AMVP mode can be applied to next piece.In the feelings to next piece of application AMVP mode Under condition, encoding device can generate motion vector predictor candidate list based on the motion information of next piece of contiguous block.

Figure 10 show based on object block modified motion information update object block motion information in the case where with mesh Mark the example of next piece neighbouring of motion vector predictor candidate list of block.Encoding device can be included in motion-vector prediction Motion information is selected according to specified conditions in motion information in device candidate list, and sends and indicates selected motion information Motion vector difference (MVD) value of index, the motion vector of selected motion information is the motion vector of next piece of motion information.With Above-mentioned be inserted in the block of the motion information of neighbouring next piece of the update of instruction under merging patterns merges space in candidate list Or the method that the time merges in candidate blocks is similar, the process for generating the list of the motion information of the contiguous block including AMVP mode can It is applied to using the motion information of the update of contiguous block as space or the side of the candidate insertion of motion vector prediction device (MVP) Method.As shown in Figure 10, encoding device can detect neighbouring next piece of object block A0 and object block A1 according to the order of arrow direction Update motion information whether according to predefined specified conditions, and derive first it is detected that according to specified conditions more New motion information is as including the MVP A in motion predictor candidate list.

In addition, encoding device can be detected according to the order of arrow direction neighbouring next piece object block B0, object block B1 and The motion information of the update of object block B2 whether according to predefined specified conditions, and derive first it is detected that according to specific item The motion information of the update of part is as including the MVP B in motion predictor candidate list.

In addition, next piece can be located in the picture different from object block, and next piece of time MVP candidate may include In motion vector predictor candidate list.As shown in Figure 10, encoding device can be according to the mesh in next piece of reference pictures The order for marking the motion information of the update of the motion information of the update of block T0 to object block T1 is detected whether according to predefined Specified conditions, and derive first it is detected that the update according to specified conditions motion information as be included in motion predictor time Select the MVP Col in list.It by being replaced by modified motion information but is updated in MVP A, MVP B and/or MVP Col In the case where motion information, the motion vector predictor that the motion information of update can be used for next piece is candidate.

In addition, the quantity of the MVP candidate in the motion predictor candidate list derived is less than the case where specific quantity Under, encoding device derives zero vector and as MVP zero and is included into motion predictor candidate list.

Figure 11 show based on object block modified motion information update object block motion information in the case where with mesh Mark the example of next piece neighbouring of motion vector predictor candidate list of block.It includes modified for being in the motion information of object block In the case where the motion information of the update of motion information and existing motion information, encoding device can be examined according to the order of arrow direction The existing motion information of neighbouring next piece of object block A0 and object block A1 are surveyed whether according to specified conditions, and derives and detects first To the existing motion information according to specified conditions as including the MVP A in motion predictor candidate list.

In addition, the modified movement that encoding device can detect object block A0 and object block A1 according to the order of arrow direction is believed Breath whether according to specified conditions, and derive first it is detected that the modified motion information according to specified conditions as being included in fortune The MVP A of update in dynamic fallout predictor candidate list.

In addition, encoding device can be detected according to the order of arrow direction neighbouring next piece object block B0, object block B1 and The existing motion information of object block B2 whether according to predefined specified conditions, and derive first it is detected that according to specified conditions Existing motion information as including the MVP B in motion predictor candidate list.

In addition, encoding device can detect repairing for object block B0, object block B1 and object block B2 according to the order of arrow direction Positive motion information whether according to specified conditions, and derive first it is detected that the motion information of the update according to specified conditions make It is the MVP B for including update in motion predictor candidate list.

In addition, encoding device can be according to the existing motion information of the object block T0 in next piece of reference pictures to mesh The order for marking the existing motion information of block T1 detects whether according to specified conditions, and derive first it is detected that according to specified conditions Existing motion information as including the MVP Col in motion predictor candidate list.

In addition, encoding device can be arrived according to the modified motion information of the object block T0 in next piece of reference pictures The order of the modified motion information of object block T1 detects whether according to specified conditions, and derive first it is detected that according to specific The modified motion information of condition is as including the MVP Col in motion predictor candidate list.

Figure 12 schematically shows the method for video coding of encoding device according to the present invention.Method shown in Figure 12 can be by Encoding device shown in FIG. 1 executes.In detailed example, the step S1200 to S1230 of Figure 12 can be by the prediction list of encoding device Member executes, and step S1240 can be executed by the memory of encoding device.

The motion information (step S1200) of encoding device generation object block.Encoding device can be pre- to object block application interframe It surveys.To object block application inter-prediction, encoding device can be by using skip mode, merging patterns and adaptive At least one of motion-vector prediction (AMVP) mode generates the motion information of object block.Motion information is referred to alternatively as first Motion information.In the case where skip mode and merging patterns, encoding device can be based on the motion information of the contiguous block of object block To generate the motion information of object block.Motion information may include motion vector and reference picture indices.Motion information can be double Predicted motion information or single prediction motion information.Double prediction motion information may include L0 reference picture indices and L0 motion vector with And L1 reference picture indices and L1 motion vector, one-way movement information may include L0 reference picture indices and L0 motion vector or L1 Reference picture indices and L1 motion vector.L0 indicates reference pictures list L0 (list 0), and L1 indicates reference pictures list L1 (column Table 1).

In the case where AMVP mode, encoding device can be used the motion vector of the contiguous block of object block as motion vector Fallout predictor (MVP) carrys out the motion vector of derived object block, and generates the reference picture indices including motion vector and motion vector Motion information.

Encoding device is by deriving forecast sample (step based on the inter-prediction of motion information performance objective block S1210).Encoding device can generate the pre- of object block based on including reference picture indices in motion information and motion vector Test sample sheet.

Encoding device generates reconstructed blocks (step S1220) based on forecast sample.Encoding device can based on forecast sample come The reconstructed blocks of object block are generated, or generates the residual signals of object block and target is generated based on residual signals and forecast sample The reconstructed blocks of block.

Encoding device generates the modified motion information (step S1230) of object block based on reconstructed blocks.Encoding device can Calculate the modified reference picture indices of the particular reference picture for indicating modified motion information and specific by various methods The modified motion vector of reference pictures.Modified motion information is referred to alternatively as the second motion information.Including such as light stream (OF) The direct method and singular point matching process of method, block matching method, frequency domain method etc. use method of statistical property etc. At least one method of indirect method can be applied to this method.In addition, direct method and indirect method can be applied simultaneously.

For example, encoding device can generate modified motion information by block matching method.In this case, coding is set It is standby can be by measuring distortion level according to the difference between the reconstructed blocks of object block and the accumulation of phase sample of reference block and will These are used as cost function, it is then detected that the special datum block of reconstructed blocks.Encoding device can be based on special datum block detected To generate modified motion information.In other words, encoding device produces modified motion information comprising instruction special datum rope The modified motion vector of the modified reference picture indices and the special datum block in instruction particular reference picture drawn.Coding Equipment can detect the difference according to the phases of the reconstructed blocks of object block between samples in the reference block in particular reference picture The smallest reference block of summation of absolute value (or square value) repaired as special datum block, and based on the special datum block to derive Positive motion information.As the method for the summation for indicating absolute value of the difference, absolute difference and (SAD) can be used.In this case, Above-mentioned formula 1 can be used to calculate the summation of absolute value of the difference.In addition, the method as the summation for indicating absolute value of the difference, can make With mean square error (MSE).In this case, above-mentioned formula 2 can be used to calculate the summation of absolute value of the difference.

In addition, the particular reference picture of modified motion information can be derived as including the base in (first) motion information Reference pictures indicated by quasi- picture index, and the region of search for being used to detect special datum block can be limited to be located at away from base The reference block that included motion vector related with reference pictures derives in reference pictures in (first) motion information is Reference block in presumptive area.It is located at that is, encoding device can derive away from based on including in (first) motion information and benchmark The reference block that the related motion vector of picture derives in reference pictures is in the reference block in presumptive area and reconstructed blocks The smallest reference block of SAD as special datum block.

In addition, reconstructed blocks can be divided into multiple sub- reconstructed blocks in the case where the size of reconstructed blocks is greater than predefined size, And it can be in particular reference picture to derive specific sub- reconstructed blocks as unit of sub- reconstructed blocks.In this case, encoding device Special datum block can be derived based on the specific sub- reconstructed blocks derived.

As another example, encoding device can generate modified motion information by OF method.In this case, it compiles Decoding apparatus can calculate the motion vector of the modified motion information of object block based on such hypothesis: object in object block Speed is uniform, and indicates that the sample value of the sample of the object does not change in the picture.Movement arrow can be calculated by above-mentioned formula 6 Amount.Instruction includes the region that the region of the sample of the object in object block can be configured to 3 × 3 sizes.

In the case where calculating modified motion information by the above method, encoding device can calculate modified motion information To have format identical with (first) motion information.That is, encoding device can calculate modified motion information to transport in double prediction There is format identical with motion information between dynamic information and single prediction motion information.Double prediction motion information may include L0 benchmark Picture index and L0 motion vector and L1 reference picture indices and L1 motion vector, one-way movement information may include that L0 benchmark is drawn Face index and L0 motion vector or L1 reference picture indices and L1 motion vector.L0 indicates reference pictures list L0 (list 0), L1 It indicates reference pictures list L1 (list 1).

In addition, double prediction can be used in the target picture for including object block executing, encoding device can pass through the above method Modified motion information is calculated as double prediction motion information.However, being calculated as double prediction fortune in modified motion information After dynamic information, in the case where one of special datum block derived by double prediction motion information above blocks, in addition to packet Except the motion information for including the reference pictures list of the particular reference picture of the special datum block blocked, encoding device can be incited somebody to action Modified motion information is derived as single prediction motion information.When according to the reference block and target derived based on modified motion information When difference between the sample of the phase of the reconstructed blocks of block is greater than specific threshold, it may be determined whether block.Threshold value can be preparatory Configuration.

For example, modified motion information is calculated as double prediction motion information by encoding device, and according to based on amendment Motion information L0 motion vector and L0 reference picture indices derive special datum block and object block reconstructed blocks phase Sample between difference be greater than preconfigured threshold value in the case where, modified motion information can be derived as wrapping by encoding device Include the single prediction motion information of L1 motion vector and L1 reference picture indices.

For another example, modified motion information is calculated as double prediction motion information by encoding device, and according to based on amendment Motion information L1 motion vector and L1 reference picture indices derive special datum block and object block reconstructed blocks phase Sample between difference be greater than preconfigured threshold value in the case where, modified motion information can be derived as wrapping by encoding device Include the single prediction motion information of L0 motion vector and L0 reference picture indices.

In addition, according to the weight based on L0 motion vector and L0 reference picture indices the special datum block derived and object block In the case that difference between the sample of the phase of building block is greater than preconfigured threshold value and according to based on L1 motion vector Difference between the sample of the phase of the reconstructed blocks of the special datum block and object block of L1 reference picture indices derivation is greater than pre- In the case where the threshold value first configured, encoding device can the motion information of contiguous block of derived object block believe as modified movement Breath, or modified motion information can not be calculated.

In addition, encoding device can be optionally comprised in the modified reference pictures in modified motion information by various methods The indicated particular reference picture of index.

For example, encoding device can include the base for selecting to encode recently in the reference pictures in reference pictures list L0 Quasi- picture and the modified L0 reference picture indices for generating instruction reference pictures.In addition, encoding device can be drawn being included in benchmark Reference pictures in face list L1 select the reference pictures encoded recently in the middle and generate the modified L1 base of instruction reference pictures Quasi- picture index.

For another example, encoding device can include selecting picture order count in the reference pictures in L0 reference picture indices (POC) with the smallest reference pictures of the absolute value of the difference of the POC of target picture, and generate instruction reference pictures modified L0 base Quasi- picture index.In addition, encoding device can include selecting picture order in the reference pictures in L1 reference picture indices The smallest reference pictures of absolute value of the difference of the POC of (POC) and target picture are counted, and generate the modified of instruction reference pictures L1 reference picture indices.

For another example, encoding device can include selecting to belong to layering knot in the reference pictures in each reference pictures list The reference pictures of lowermost layer on structure, and generate the modified reference picture indices of instruction reference pictures.Belong to the base of lowermost layer Quasi- picture can be I slice or the reference pictures by application low quantization parameter (QP) coding.

For another example, the base of the highest reference block of reliability in reference pictures list including motion compensation may be selected in encoding device Quasi- picture, and generate the modified reference picture indices of instruction reference pictures.In other words, encoding device can be based on being included in benchmark Reference pictures in picture list carry out the special datum block of the reconstructed blocks of derived object block, and generating instruction includes derived spy Determine the modified reference picture indices of the particular reference picture of reference block.

The method of the modified reference picture indices of the above-mentioned modified motion information of generation can independently apply or these Method can combined application.

In addition, encoding device can generate the modified movement of object block based on the original block of object block although being not shown Information.Encoding device can include the special datum block of original block being derived in reference block in reference pictures, and generate and refer to Show the modified motion information of derived special datum block.

Encoding device updates the motion information (step S1240) of object block based on modified motion information.Encoding device Modified motion information can be stored and update the motion information of object block.Encoding device can be by that will be used to predict the fortune of object block Dynamic information replaces with modified motion information to update the motion information of object block.In addition, encoding device can be used for by storage All motion informations and modified motion information for predicting object block update the motion information of object block.The motion information of update It can be used for next piece of the motion information neighbouring with object block.For example, merging patterns are applied to next piece neighbouring with object block, Next piece of merging candidate list may include object block.Modified by replacing with the motion information for being used to predict object block In the case where motion information of the motion information to store object block, next piece of merging candidate list may include indicating modified fortune The object block of dynamic information.In addition, for predicting that all motion informations of object block and modified motion information are stored in target In the motion information of block, next piece of merging candidate list may include the object block indicated for predicting the motion information of object block And the object block of the modified motion information of instruction.Indicate that the object block of modified motion information can be used as spatial neighbor candidate blocks It is inserted in and merges in candidate list, or be inserted in and merge in candidate list as time neighbor candidate block.

For another example, apply AMVP mode to object block neighbouring next piece, be similar to it is above-mentioned will be under under merging patterns The motion information of the update of one piece of neighbouring object block, which is inserted in, to be merged in candidate list as space or time neighbouring movement letter The method of breath can be inserted in the candidate of the motion vector predictor in above-mentioned next piece using by the motion information of the update of contiguous block As space or the method for motion vector prediction device candidate in list.That is, encoding device produce include and next piece of neighbour The motion vector predictor candidate list of the motion information of the update of close object block.

For example, the motion information in the update with next piece of neighbouring object block only includes the case where modified motion information Under, motion vector predictor candidate list may include that modified motion information is candidate as spatial motion vector prediction device.

For another example, in the modified motion information that the motion information of the update with next piece of neighbouring object block includes object block In the case where existing motion information, motion vector predictor candidate list may include showing with next piece of neighbouring object block Have in motion information according to the existing motion information of specified conditions selection and in the amendment with next piece of neighbouring object block Motion information according to specified conditions select modified motion information, as additional space motion vector predictor wait Choosing.

In addition, next piece of motion vector predictor candidate list may include in next piece of reference pictures with next piece The motion information of the update of the same position block of position same position and with motion vector prediction device candidate list same position The motion information of the update of the contiguous block of block.For example, the movement of the update in the same position block of the position same position with next piece The motion information of the update of the contiguous block of information and same position block only includes the case where each piece of modified motion information Under, motion vector predictor candidate list may include modified motion information as motion vector prediction device candidate list.

For another example, the motion information of update include same position block, same position block contiguous block corresponding modified fortune In the case where dynamic information and the existing motion information of same position block, motion vector predictor candidate list can respectively include repairing According to the modified motion information of specified conditions selection and existing motion information in the middle according to specific in positive motion information The existing motion information of condition selection, it is candidate as individual spatial motion vector prediction device.

In addition, the motion information stored and the motion information being used for transmission can have different resolution ratio.For example, coding Equipment is encoded and is sent to the information for motion information, and is corrected for storing with the neighbouring motion information of object block Motion information, the unit including the motion vector in motion information can have the unit of 1/4 part sample, modified movement The unit of the motion vector of information can indicate the unit of 1/8 part sample or 1/16 part sample.

In addition, although being not shown, encoding device can be executed by the reconstructed blocks based on object block (first) motion information with Comparison procedure between modified motion information determines whether to update (first) motion information of object block.For example, coding is set It is standby can be by will be based on the data of the residual signals of the reconstructed blocks of the modified motion information special datum block derived and object block Amount is compared to determine whether to update with the data volume of the reference block derived and the residual signals of reconstructed blocks based on motion information Object block.In these data volumes, in the lesser situation of data volume of special datum block and the residual signals of reconstructed blocks, compile Decoding apparatus, which can determine, updates motion information and object block.In addition, in these data volumes, in special datum block and reconstructed blocks In the no small situation of the data volume of residual signals, encoding device, which can determine, does not update motion information and object block.

In addition, encoding device can be executed between motion information and modified motion information by the original block based on object block Comparison procedure come determine whether update object block (first) motion information.For example, encoding device can be by that will be based on correcting The motion information special datum block derived and the data volume of residual signals of original block of object block pushed away with based on motion information The data volume of the residual signals of the reference block and original block led is compared to determine whether to update object block.In these data volumes In the middle, in the lesser situation of data volume of special datum block and the residual signals of original block, encoding device can determine update fortune Dynamic information and object block.In addition, in these data volumes, the data volume of special datum block and the residual signals of original block not In the case where small, encoding device, which can determine, does not update motion information and object block.

In addition, encoding device produces and encodes the additional information for indicating whether to update and exports it by bit stream. For example, indicating whether that the additional information updated is referred to alternatively as updating mark.Updating the case where block is 1 can indicate motion information by more Newly, updating the case where block is 0 can indicate that motion information does not update.For example, updating mark can be sent as unit of PU.Alternatively, Updating mark can send as unit of CU, as unit of CTU or as unit of slice, and can pass through such as parameter sets (PPS) higher level of the unit of unit or sequence parameter set (SPS) is sent.

In addition, encoding device, which also produces, indicates that the movement of the difference between existing motion vector and modified motion information is sweared Amount updates poor information and is encoded and exported.Motion vector, which updates poor information, to be sent as unit of PU.

Although being not shown, encoding device can will be about the information coding of the residual sample of object block and output.About residual error The information of sample may include the transformation coefficient of residual sample.

Figure 13 schematically shows the video encoding/decoding method of decoding device according to the present invention.Method shown in Figure 13 can be by Decoding device shown in Fig. 2 executes.In detailed example, the step S1300 of Figure 13 can be held by the entropy decoding unit of decoding device Row, step S1310 to S1340 can be executed by the predicting unit of decoding device, and step S1350 can be held by the memory of decoding device Row.

Decoding device obtains the information (step S1300) of the inter-prediction for object block by bit stream.It can be to target Block application inter prediction or infra-frame prediction.To object block application inter-prediction, decoding device can pass through bit stream Obtain the information of the inter-prediction for object block.In addition, decoding device can obtain the existing of instruction object block by bit stream The motion vector of difference between motion vector and modified motion information updates poor information.In addition, decoding device can pass through bit Stream obtains the additional information for whether updating object block.For example, indicating whether that the additional information updated is referred to alternatively as updating mark.

Decoding device is based on the information about inter-prediction come the motion information (step S1310) of derived object block.Movement Information is referred to alternatively as the first motion information.Information about inter-prediction can indicate skip mode, merging patterns and adaptive meet the tendency of It is applied to the mode of object block in dynamic vector prediction (AMVP) mode.It is applied to object block in skip mode or merging patterns In the case of, decoding device produces the merging candidate list of the contiguous block including object block, and obtains instruction and be included in merging time The merging of the contiguous block in the contiguous block in list is selected to index.Merging index may include in the information about inter-prediction. Decoding device can derive motion information of the motion information for merging the indicated contiguous block of index as object block.

In the case where AMVP mode is applied to object block, merging patterns are similar to, decoding device can be based on object block Contiguous block generates list.It includes the contiguous block in the contiguous block in list generated that decoding device, which produces instruction, Motion vector difference (MVD) indicated by index and index between the motion vector of contiguous block and the motion vector of object block. Index and MVD can be included in the information about inter-prediction.Decoding device can be based on the fortune of the indicated contiguous block of index Dynamic vector and MVD generate the motion information of object block.

Motion information may include motion vector and reference picture indices.Motion information can be double prediction motion information or list Predicted motion information.Double prediction motion information may include L0 reference picture indices and L0 motion vector and L1 reference picture indices With L1 motion vector, one-way movement information may include L0 reference picture indices and L0 motion vector or L1 reference picture indices and L1 Motion vector.L0 indicates reference pictures list L0 (list 0), and L1 indicates reference pictures list L1 (list 1).

Decoding device derives forecast sample (step by executing the inter-prediction to object block based on motion information S1320).Decoding device can generate the pre- of object block based on including reference picture indices in motion information and motion vector Test sample sheet.

Decoding device generates reconstructed blocks (step S1330) based on forecast sample.To object block application skip mode In the case of, decoding device can generate the reconstructed blocks of object block based on forecast sample.To object block application merging patterns or In the case where AMVP mode, decoding device can be generated the residual signals of object block by bit stream and based on residual signals and in advance Test sample generated the reconstructed blocks of object block originally.

Encoding device is based on reconstructed blocks come the modified motion information (step S1340) of derived object block.Decoding device can Calculate the modified reference picture indices of the particular reference picture for indicating modified motion information and specific by various methods The modified motion vector of reference pictures.Modified motion information is referred to alternatively as the second motion information.Including such as light stream (OF) The direct method and singular point matching process of method, block matching method, frequency domain method etc. use method of statistical property etc. At least one method of indirect method can be applied to this method.In addition, direct method and indirect method can be applied simultaneously.

For example, decoding device can generate modified motion information by block matching method.In this case, decoding is set It is standby distortion level to be measured simultaneously according to the difference between the reconstructed blocks of object block and the sample of the phase of reference block by accumulating These are used as cost function, it is then detected that the special datum block of reconstructed blocks.Decoding device can be based on special datum detected Block generates modified motion information.In other words, decoding device produces modified motion information comprising instruction special datum The modified reference picture indices of index and the modified motion vector of the special datum block in instruction particular reference picture.Solution Decoding apparatus can be detected in the reference block in particular reference picture between the sample of the phase of the reconstructed blocks according to object block The smallest reference block of summation of absolute value of the difference (or square value) is derived as special datum block, and based on the special datum block Modified motion information.As the method for the summation for indicating absolute value of the difference, absolute difference and (SAD) can be used.In such case Under, above-mentioned formula 1 can be used to calculate for the summation of absolute value of the difference.In addition, the method as the summation for indicating absolute value of the difference, it can It uses mean square error (MSE).In this case, above-mentioned formula 2 can be used to calculate for the summation of absolute value of the difference.

In addition, the particular reference picture of modified motion information can be derived as by including in (first) motion information The region of search for being used to detect special datum block can be limited to be located at away from based on packet by the reference pictures of reference picture indices instruction Including the reference block that the motion vector related with reference pictures in (first) motion information derives in particular reference picture is Reference block in presumptive area.It is located at that is, decoding device can derive away from based on including having in motion information with reference pictures The reference block that the motion vector of pass derives in particular reference picture is in the reference block in presumptive area with reconstructed blocks The smallest reference block of SAD is as special datum block.

In addition, reconstructed blocks can be divided into multiple sub- reconstructed blocks in the case where the size of reconstructed blocks is greater than predefined size, It can be in particular reference picture by sub- reconstructed blocks specific as unit of sub- reconstructed blocks.In this case, decoding device can be based on institute The specific sub- reconstructed blocks derived derive special datum block.

For another example, decoding device can generate modified motion information by OF method.In this case, decoding device can The modified motion vector of the modified motion information of object block: the speed of the object in object block is calculated based on such hypothesis Degree uniformly, and indicates that the sample value of the sample of the object does not change in the picture.Motion vector can be counted by above-mentioned formula 6 It calculates.Instruction includes the region that the region of the sample of the object in object block can be configured to 3 × 3 sizes.

In the case where calculating modified motion information by the above method, decoding device can calculate modified motion information To have format identical with (first) motion information.That is, decoding device can calculate modified motion information to transport in double prediction There is format identical with motion information between dynamic information and single prediction motion information.Double prediction motion information may include L0 benchmark Picture index and L0 motion vector and L1 reference picture indices and L1 motion vector, one-way movement information may include that L0 benchmark is drawn Face index and L0 motion vector or L1 reference picture indices and L1 motion vector.L0 indicates reference pictures list L0 (list 0), L1 It indicates reference pictures list L1 (list 1).

In addition, double prediction can be used in the target picture for including object block executing, decoding device can pass through the above method Modified motion information is calculated as double prediction motion information.However, being calculated as double prediction movement in modified motion information After information, in the case where one of reference block derived by double prediction motion information above blocks, in addition to including occurring Except the motion information of the reference pictures list of the particular reference picture of the special datum block blocked, decoding device can will be modified Motion information is derived as single prediction motion information.When according to the special datum block and object block derived based on modified motion information Reconstructed blocks phase sample between difference be greater than specific threshold when, if blocking can be confirmed as occurring.Threshold value It can be pre-configured with.

For example, modified motion information is calculated as double prediction motion information by decoding device, and according to based on amendment Motion information L0 motion vector and L0 reference picture indices derive special datum block and object block reconstructed blocks phase Sample between difference be greater than preconfigured threshold value in the case where, modified motion information can be derived as wrapping by encoding device Include the single prediction motion information of L1 motion vector and L1 reference picture indices.

For another example, modified motion information is calculated as double prediction motion information by decoding device, and according to based on amendment Motion information L1 motion vector and L1 reference picture indices derive special datum block and object block reconstructed blocks phase Sample between difference be greater than preconfigured threshold value in the case where, modified motion information can be derived as wrapping by decoding device Include the single prediction motion information of L0 motion vector and L0 reference picture indices.

In addition, according to the weight based on L0 motion vector and L0 reference picture indices the special datum block derived and object block In the case that difference between the sample of the phase of building block is greater than preconfigured threshold value and according to based on L1 motion vector Difference between the sample of the phase of the reconstructed blocks of the special datum block and object block of L1 reference picture indices derivation is greater than pre- In the case where the threshold value first configured, decoding device can the motion information of contiguous block of derived object block believe as modified movement Breath, or modified motion information can not be calculated.

In addition, decoding device can be optionally comprised in the picture of the modified benchmark in modified motion information by various methods The indicated particular reference picture of face index.

For example, decoding device can include the base for selecting to encode recently in the reference pictures in reference pictures list L0 Quasi- picture, and generate the modified L0 reference picture indices of instruction reference pictures.In addition, decoding device can be drawn being included in benchmark Reference pictures in face list L1 select the reference pictures encoded recently in the middle, and generate the modified L1 base of instruction reference pictures Quasi- picture index.

For another example, decoding device can include selecting picture order count in the reference pictures in L0 reference picture indices (POC) with the smallest reference pictures of the absolute value of the difference of the POC of target picture, and generate instruction reference pictures modified L0 base Quasi- picture index.In addition, decoding device can include selecting picture order in the reference pictures in L1 reference picture indices The smallest reference pictures of absolute value of the difference of the POC of (POC) and target picture are counted, and generate the modified of instruction reference pictures L1 reference picture indices.

For another example, decoding device can include selecting to belong to layering knot in the reference pictures in each reference pictures list The reference pictures of lowermost layer on structure, and generate the modified reference picture indices of instruction reference pictures.Belong to the base of lowermost layer The reference pictures that quasi- picture can be I slice or be encoded by application low quantization parameter (QP).

For another example, decoding device can select the highest reference block of reliability including motion compensation in reference pictures list Reference pictures, and generate the modified reference picture indices of instruction reference pictures.In other words, decoding device can be based on being included in base Reference pictures in quasi- picture list carry out the special datum block of the reconstructed blocks of derived object block, and generating instruction includes being derived The modified reference picture indices of the particular reference picture of special datum block.

In addition, decoding device can obtain the existing motion vector of instruction object block by bit stream and modified movement is sweared The motion vector of difference between amount updates poor information.In this case, decoding device can be transported by (first) to object block Dynamic information and motion vector update poor information summation to derive modified motion information.Motion vector updates poor information can be above-mentioned PU is unit transmission.

Decoding device updates the motion information (step S1350) of object block based on modified motion information.Decoding device Modified motion information can be stored and update the motion information of object block.Decoding device can be by that will be used to predict the fortune of object block Dynamic information replaces with modified motion information to update the motion information of object block.In addition, decoding device can be used for by storage All motion informations and modified motion information for predicting object block update the motion information of object block.The motion information of update It can be used for next piece of the motion information neighbouring with object block.

For example, applying merging patterns to next piece neighbouring with object block, next piece of merging candidate list may include mesh Mark block.The movement of object block letter is being stored by the way that the motion information for being used to predict object block is replaced with modified motion information In the case where breath, next piece of merging candidate list may include the object block for indicating modified motion information.In addition, for predicting All motion informations and modified motion information of object block are stored in the motion information of object block, and next piece of merging is waited Selecting list may include indicating the target of object block and the modified motion information of instruction for predicting the motion information of object block Block.It indicates that the object block of modified motion information can be used as spatial neighbor candidate blocks and be inserted in merge in candidate list, Huo Zhezuo It is inserted in and merges in candidate list for time neighbor candidate block.

For another example, apply AMVP mode to object block neighbouring next piece, be similar to it is above-mentioned will be under under merging patterns The motion information of the update of one piece of neighbouring object block, which is inserted in, to be merged in candidate list as space or time neighbouring movement letter The method of breath can be inserted in the candidate of the motion vector predictor in above-mentioned next piece using by the motion information of the update of contiguous block As space or the method for motion vector prediction device candidate in list.That is, decoding device produce include and next piece of neighbour The motion vector predictor candidate list of the motion information of the update of close object block.

For another example, in the modified motion information that the motion information of the update with next piece of neighbouring object block includes object block In the case where existing motion information, motion vector predictor candidate list may include the amendment with next piece of neighbouring object block Motion information according to the modified motion information of specified conditions selection and existing with next piece of neighbouring object block The existing motion information selected in motion information according to specified conditions, it is candidate as additional space motion vector predictor.

For another example, the motion information of update include same position block, same position block contiguous block corresponding modified fortune In the case where dynamic information and the existing motion information of same position block, motion vector predictor candidate list can respectively include repairing According to the modified motion information of specified conditions selection and existing motion information in the middle according to specific in positive motion information The existing motion information that condition selects is candidate as individual spatial motion vector prediction device.

In addition, the motion information stored and the motion information being used for transmission can have different resolution ratio.For example, decoding Equipment decodes and sends the information for motion information, and stores modified movement for the neighbouring motion information of object block Information, the unit including the motion vector in motion information can have the unit of 1/4 part sample, modified motion information The unit of motion vector can indicate the unit of 1/8 part sample or 1/16 part sample.

In addition, although being not shown, decoding device can be executed by the reconstructed blocks based on object block (first) motion information and Comparison procedure between modified motion information determines whether to update (first) motion information of object block.For example, decoding is set It is standby can be by will be based on the data of the residual signals of the reconstructed blocks of the modified motion information special datum block derived and object block Amount is compared to determine whether to update with the data volume of the reference block derived and the residual signals of reconstructed blocks based on motion information Object block.In these data volumes, in the lesser situation of data volume of special datum block and the residual signals of reconstructed blocks, solution Decoding apparatus, which can determine, updates motion information and object block.In addition, in these data volumes, in special datum block and reconstructed blocks In the no small situation of the data volume of residual signals, decoding device, which can determine, does not update motion information and object block.

In addition, decoding device can obtain the additional information for indicating whether to update object block by bit stream.For example, instruction The additional information whether updated is referred to alternatively as updating mark.Updating the case where block is 1 can indicate that motion information is updated, and update block The case where being 0, can indicate that motion information does not update.For example, updating mark can be sent as unit of PU.Alternatively, mark is updated It can be sent as unit of CU, as unit of CTU or as unit of slice, and the list of such as parameter sets (PPS) can be passed through The higher level of the unit of position or sequence parameter set (SPS) is sent.

According to aforementioned present invention, after the decoding process of object block, the modified motion information of object block is calculated, and It can be updated to more accurate motion information, overall coding efficiency can improve as a result,.

In addition, according to the present invention, can be derived based on the motion information of the update of object block and neighbouring next of object block The motion information of block, and the propagation of distortion can be reduced, overall coding efficiency can improve as a result,.

In the above-described embodiment, method is described as by series of steps or box based on flow chart.However, the disclosure is not It is limited to these order of steps.Some steps can be carried out simultaneously or be carried out according to the order different from above-mentioned steps order.In addition, It will be understood by those skilled in the art that step shown in the flowchart is not exclusive.It will be understood that in the model for not influencing the disclosure In the case where enclosing, it may include other steps can delete one or more steps in flow chart.

It can be realized in software according to above-mentioned disclosed method.According to the code device and/or decoding apparatus of the disclosure It may include in the device that for example, TV, computer, smart phone, set-top box or display device execute image procossing.

When embodiment of the present disclosure is implemented in software, the above method can by execute above-mentioned function module (process, Function etc.) Lai Shixian.These modules can be stored in memory and be executed by processor.Memory can be in the inside of processor Or it is external, and various well known means can be used to be connected to processor for memory.Processor may include specific integrated circuit (ASIC), other chipsets, logic circuit and/or data processing equipment.Memory may include ROM (read-only memory), RAM (random access memory), flash memory, storage card, storage medium and/or other storage devices.

Claims

1. it is a kind of by decoding device execute picture decoding method, the picture decoding method the following steps are included:

The information of the inter-prediction about object block is obtained by bit stream；

Based on the motion information for deriving the object block about the information of inter-prediction；

Forecast sample is derived by executing the inter-prediction for the object block based on the motion information；

Reconstructed blocks are generated based on the forecast sample；

The modified motion information of the object block is derived based on the reconstructed blocks；And

The motion information of the object block is updated based on the modified motion information.

2. picture decoding method according to claim 1, the picture decoding method is further comprising the steps of:

Determine whether to update the motion information of the object block,

Wherein, when determining the motion information for updating the object block, with the amendment derived based on the reconstructed blocks Motion information based on update the motion information of the object block.

3. picture decoding method according to claim 1, wherein the motion information of update is for deriving and the object block Next piece neighbouring of motion information.

4. picture decoding method according to claim 1, wherein when including the motion vector in the motion information When unit indicates the unit of 1/4 part sample, the unit including the motion vector in the modified motion information indicates 1/8 The unit of part sample.

5. picture decoding method according to claim 1, wherein derive the modified motion information the step of include with Lower step:

The particular reference picture of the modified motion information is derived in reference pictures；

Special datum block is detected in the reference block in the particular reference picture；And

Derive includes indicating the modified reference picture indices of the particular reference picture and repairing for the instruction special datum block The modified motion information of positive motion vector.

6. picture decoding method according to claim 5, wherein the special datum block is derived as the special datum In reference block in picture with the reconstructed blocks difference and the smallest reference block of SAD, wherein the SAD based on following formula come It determines:

Here, Block_cur(i, j) indicates the reconstructed sample of (i, j) coordinate in the reconstructed blocks, Block_ref(i, j) indicates institute The reconstructed sample of (i, j) coordinate in reference block is stated, " width " indicates the width of the reconstructed blocks, and " height " is indicated The height of the reconstructed blocks.

7. picture decoding method according to claim 5, wherein the special datum of the modified motion information is drawn Face be derived as by include in the motion information reference picture indices instruction reference pictures, and

Wherein, it is included within away from based on including that the motion vector in the motion information is pushed away from the particular reference picture The reference block led is to be derived as the spy with the smallest reference blocks of SAD of the reconstructed blocks in the reference block in presumptive area Determine reference block.

8. picture decoding method according to claim 5, wherein the step of deriving the special datum block includes following step It is rapid:

When the size of the reconstructed blocks is greater than preconfigured size, the reconstructed blocks are divided into multiple sub- reconstructed blocks；

To derive specific subbase as unit of each sub- reconstructed blocks in the particular reference picture of the modified motion information Quasi- block；And

The special datum block is derived based on the specific sub- reference block derived.

9. picture decoding method according to claim 1, wherein by based on including that L0 in the motion information is transported The light stream OF method that dynamic vector and L1 motion vector execute derives the modified motion information of the object block.

10. picture decoding method according to claim 1, wherein the modified motion information is double prediction movement letter Breath, and

Wherein, the double prediction motion information includes L0 motion vector and L0 reference picture indices and L1 motion vector and L1 base Quasi- picture index.

11. picture decoding method according to claim 1 is sweared when being moved according to the L0 based on the modified motion information Measure L0 reference picture indices derive special datum block and the object block the reconstructed blocks between phase sample it Between difference be greater than preconfigured threshold value when, the modified motion information is derived as including L1 motion vector and L1 benchmark The single prediction motion information of picture index.

12. picture decoding method according to claim 10, wherein the L0 benchmark of the modified motion information is drawn Face indexes the picture order count POC indicated include in the reference pictures in reference pictures list L0 and includes the target The smallest reference pictures of the absolute value of the difference of the POC of the target picture of block, and

Wherein, the L1 reference picture indices instruction of the modified motion information is included in the reference pictures list L1 Reference pictures in picture order count POC with include the object block the target picture POC difference it is absolute It is worth the smallest reference pictures.

13. picture decoding method according to claim 1, wherein described in being updated based on the modified motion information The motion information of object block and when to neighbouring next piece application merging patterns of the object block, described next piece Merging candidate list includes the object block, and

Wherein, the motion information of the update of the object block includes the modified motion information.

14. a kind of decoding device for executing image decoding, the decoding device include:

Entropy decoder, the entropy decoder are configured as obtaining the information of the inter-prediction about object block by bit stream；

Fallout predictor, the fallout predictor are configured as based on the movement letter for deriving the object block about the information of inter-prediction Breath derives forecast sample by executing the inter-prediction for the object block based on the motion information, based on described pre- Test sample generated the reconstructed blocks of the object block originally, and the modified movement of the object block is derived based on the reconstructed blocks Information；And

Memory, the memory are configured as updating the movement letter of the object block based on the modified motion information Breath.

15. decoding device according to claim 14, wherein the fallout predictor determines described repair based on the reconstructed blocks Whether positive motion information is motion information more more accurate than motion information based on the information inference about inter-prediction, And

When the modified motion information is more more accurate than the motion information based on the information inference about inter-prediction When motion information, the memory updates the motion information of the object block based on the modified motion information.